SKILLSHUB / AGENT WORKFLOWS

Mastra FactoryFrom an issue to a reviewed PR.

Explained through the decisions your team still owns.

An open-source delivery environment connecting issue intake, persistent coding agents and repository workspaces.

One issue. Three human checkpoints.

The official lifecycle runs through intake, triage, planning, building, review and completion. Below is our illustrative CSV scenario, not a measured Factory result.

01 / TRIAGE

Agent asks what the export should contain.

You confirm: selected rows only; preserve current permissions.

02 / PLANNING

Agent proposes a change and validation plan.

You approve scope: include a populated export and an empty-selection check.

03 / REVIEW

Agent implementation leads to a pull request.

You inspect the diff and checks, then decide whether to merge.

Before you connect a repository

npm create factory

Prepare GitHub access, a model provider and a human reviewer. Linear is optional.

The default project uses Mastra Platform for authentication, database and sandboxes. Check the provider and platform costs before running it.

Command checked against the official page on September 12, 2026; not executed for this guide.

Take a clear first issue with you

Adapt this original example to a repository you control. Replace the reproduction steps and expected behavior with your own; do not treat these sample acceptance criteria as your product requirements.

View the issue and acceptance criteria
Title: CSV export contains headers but no selected rows

Reproduction: select two permitted rows and export CSV.
Observed: the file contains only a header.
Expected: the selected rows are included, with the existing columns.

Acceptance criteria:
- Two selected rows produce exactly two data rows.
- An empty selection follows the existing product behavior.
- Rows the user cannot access are never exported.
- Commas, quotes and line breaks remain valid CSV.

Scope: preserve the current permissions, columns and export UI.
Before building: identify the cause and propose the smallest fix.
Before merging: attach the relevant diff and actual check results.
Human reviewer: assign a repository maintainer.

At review, compare the exported rows with the selected IDs, inspect an empty selection and a denied-access case, and open a CSV containing punctuation. Record failures as well as passes. A created PR alone is not proof that the bug is fixed.

Is this the right starting point?

Our recommendation: consider Factory when a team needs a shared issue-to-PR workflow with human approval. For one small coding task, first check whether your existing coding assistant is sufficient.

Sources & verification

SkillsHub explains external tools. This page does not run Factory or connect your repositories. Product facts are source-checked; the walkthrough and issue template are editorial examples.

AE
AgentSkillsHub Editorial TeamAI Agent Infrastructure Reviewers

The AgentSkillsHub editorial team evaluates MCP servers, Claude skills, and AI agent integrations for security, reliability, and practical deployment readiness. Every listing undergoes permission audit, README analysis, and operational risk triage before publication.

  • Reviewed 450+ MCP server repositories
  • Developed security grading methodology (A-F)
  • Published agent deployment safety guidelines
Updated: 2026-09-12github